Mannequins Just Friends

parlej · 2024-03-07T10:05:01.906Z

Dave Brown is a treasure.

marcoluqueh · 2024-03-07T10:34:13.597Z

That’s great material! But in this case, just trying to find how the envelope acts when modified by Time and Intone. The idea is to generate via Norns+Crow the same envelope that the JF’s Vactrol model is receiving, not to recreate a vactrol.

Am I understanding the link you sent wrong?

parlej · 2024-03-07T12:26:30.839Z

Nah, the misunderstanding is all mine.

I don’t think there is published behaviour on how the env>vac interaction has been implemented… but I’m sure you could get an pretty good approximation but looking at the trace of the envelopes produced by JF as a control voltage, and combining with the rise/fall characteristics in the DB page.

Perhaps even a meshing of envelope follower & cv recording on the audio from JF into Crow could get you on the path…

There’s also a func.gen implementation (in C) tucked away in the Whimsical DSP repo - no idea if this is exactly what JF uses, but a nice nugget of intel nonetheless.

marcoluqueh · 2024-03-07T15:04:01.240Z

Oh, I think this might be helpful. Thank you!

Galapagoose · 2024-03-08T18:17:53.485Z

github.com

whimsicalraps/wrDsp/blob/wavefold/wrVtl.c

#include "wrVtl.h"

#include <stdlib.h>
#include <stdio.h> // printf

#include "wrGlobals.h"
#include "wrMath.h" // math_get_ramps()

vtl_t* vtl_init( void ){
    vtl_t* self = malloc( sizeof( vtl_t ) );
    if(self == NULL){ printf("VTL: can't malloc!\n"); }

	self->dest  = 0.0;
	self->level = 0.0;
	self->vel   = 1.0;
    vtl_mode( self, vtl_mode_sustain );
    vtl_params( self, 100.0, 0.1 );
    return self;
}

This file has been truncated. show original

this is the ‘vactrol’ emulation code in JF. it’s coupled to the wrLpGate model as well which is also non-linear, so making an exact replica is likely difficult unless you’re putting it into supercollider classes.

general idea is an integrator who’s time constant is shortened as the output level rises, plus ‘intone’ controls a step in time constant depending on whether the target level is higher or lower than the output (at 12:00, the TCs are the same).

LPG signal path can be emulated w a VCA (aka multiply) somewhere between linear & exponential response, into a 1pole lowpass filter with expo response (add a 2nd LP stage 1octave higher for more accurate emulation).

sno · 2024-03-13T10:22:32.437Z

I just upgrade the firmware of my just friends to the latest version (4.6.0), I think I was still on version (4.1.2).
Is it just me or the sound in synth mode has change to a more aggressive tone?

this is how it sounds with firmware 4.1.2
jf-firmware412.aif (3.0 MB)

this is what it sounds with firmware 4.6.0 (all knobs in the same position) it looks like the curve knob also changed its function.
jf-firmware460v01.aif (3.0 MB)

this is the closest I can get to sound similar to firmware 4.1.2, but it sounds more aggressive.
jf-firmware460v02.aif (3.0 MB)

is this the normal behavior of synth mode after upgrading to the new firmware or is there a problem with my jf?

alanza · 2024-03-13T12:38:27.973Z

i believe the curve knob now starts wavefolding as you go further CW. i didn’t listen, but my bet is yr just friends is behaving normally

3-foot-1 · 2024-03-13T13:49:43.611Z

A wave folder was added in [edit] version 4.5. the pure sine wave is now at the 12:00 position, and folding eases in as you turn CW from there.

imminent_gloom · 2024-03-13T13:54:38.988Z

In addition to the wavefolder, the square side of “shape” is sharper since the last update

sno · 2024-03-13T16:39:53.246Z

thank you so much for the insights good people!

@alanza, glad to know that my just friends is behaving normally

@3-foot-1, I kind of guessed that the pure sine wave was all knobs at 12:00 position, but it’s still very aggressive, not the gentle sines from previous firmware. but I guess it’s because of the wavefolder, right? It becomes more aggressive when TIME is further CCW.

edit: from what I can understand, if using teletype and setting the jf.note velocity to 1 I can achieve the soft sines of previous firmware. but how do I do that with Ansible Kria?

3-foot-1 · 2024-03-13T23:25:49.130Z

What you describe does not fit my mental model of how JF works, and I’m not in front of my rig, so perhaps we need to hear from someone with deeper knowledge of the 4.6 firmware.

IIUC, TIME and velocity should not influence the shape of the waveform. I wonder if you have CURVE already slightly in the wavefolding territory and it is easier to hear the folding at higher pitches and louder amplitudes. It would be interesting to look at the waveform on an oscilloscope to see whether (1) it is really a sine wave at the lower pitch and lower velocity and (2) changes shape as you adjust those parameters.

OTOH, depending on the implementation the velocity could interact with the wavefolder: if the folding happens at an absolute voltage boundary and velocity is applied pre-fold, then a low velocity would prevent folding from happening at all. If folding happens at a relative height in the waveform and velocity is applied to the folded wave, then any difference in sound quality is perceptional.

imminent_gloom · 2024-03-14T05:12:06.642Z

on on ansible, the duration setting in kria changes velocity for all the apps when jf is in synth mode.

sno · 2024-03-14T15:16:33.594Z

@3-foot-1 I look at the waveform on an oscilloscope and with all knobs at 12:00 there’s indeed a sine wave at the lower pitch and lower velocity, as well at higher pitch and velocity too. shapes are adjusted as the should when playing with the parameters.

The problem is when you play more than one voice at the same time or in round robin style, if the velocity is not set to 1, the sound becomes very aggressive. Kind like when the MIX out would distort in previous versions of the firmware. This is accentuated when the TIME knob is further to the CCW position, notes take more time to decay.
Using Ansible or Norns with Crow results in the same behavior.
I find it to be quite strange, because I use just friends a lot with pure sines and never had this issue with the firmware 4.1.2.

Thank for the duration tip for ansible @imminent_gloom! But sadly it does not fix the issue.

imminent_gloom · 2024-03-14T15:25:22.912Z

actually having listened (sorry), it sounds like there is some fm going on in both 4.6 files. stupid question: is the fm attenuator set to something other than straight up?

sno · 2024-03-14T15:45:26.785Z

no, all knobs at 12:00 minus TIME and INTONE.

imminent_gloom · 2024-03-14T15:49:29.362Z

then i have to concur wit 3foot1, this is weird. weird i ways mine has never been

Justmat · 2024-03-14T15:51:13.691Z

are you listening to the mix out? or individual outs? if you are listening to mix, could you try listening to the individual outs, and see if they’re clean? maybe it’s just the mix out overdriving

sno · 2024-03-14T15:56:44.099Z

hi @Justmat! it happens in both the mix and on individual outs going to a nearness.

Kiaaaaan · 2024-03-14T15:59:50.611Z

have the same issue, I thought the compressor on mix out changed after 4.1.2, and it seems 4.6 is better than 4.5 on this but still can’t get a clean sound at default velocity

sno · 2024-03-14T16:01:48.668Z

hi @Kiaaaaan! I thought I was alone and loosing my mind with this issue… this makes just friends useless with ansible and norns via crow for me. I can only use it with teletype or revert back to 4.1.2.